On Thu, 13 Nov 2003, Michael Holt wrote:
> On Thu, 2003-11-13 at 08:26, Bill Mullen wrote:
>
> > Including the output of "postconf -n", run on the Postfix box, might
> > be helpful also, as would the re-inclusion of the two sets of headers;
> > all that matters is the last couple of "Received:" headers from each
> > message, as those will be the ones that pertain to your sending
> > system(s).
>
> postconf -n
>
[snip]
> mydestination = $myhostname, localhost.$mydomain $mydomain
> mydomain = holt-tech.net
> myhostname = earth
> mynetworks = 192.168.0.0/24, 127.0.0.0/24
> myorigin = holt-tech.net
[snip]
Okay, I think you should at least change the "myhostname =" line, found in
the /etc/postfix/main.cf file. Having the short hostname of your Postfix
box here does you no good, as it is of utterly no use to the destination
system. OTOH, if you changed it to "holt-tech.net", then at least the name
resolves in one direction (forward), and agrees with the hostname in your
MX record for the domain. It would still fail an rDNS check, though, if
that check doesn't merely look for whether an rDNS entry exists, but goes
further to insist that it match the stated hostname (which it won't). :(
Bear in mind that the "myhostname =" setting in main.cf doesn't need to
bear even the slightest resemblance to what your system actually calls
itself; it is the string that is sent whenever Postfix identifies the
system on which it is running to other systems (both clients and servers).
As such, the value of this setting *will* have an impact on whether or not
mail is accepted from you by some servers, as it is sent in the HELO/EHLO
statement when Postfix initiates a connection as a client.
If your external hostname (the one supplied you by your ISP, and currently
"evrtwa1-ar17-4-35-151-034.evrtwa1.dsl-verizon.net") remains constant or
nearly so, then *that* is the ideal string to put into main.cf as your
"myhostname =" value, because then your name resolves in both directions.
If you can do this, it mitigates a lot of problems of this variety. The
mere fact that the hostname is obviously tied to the IP address should not
be a deal-breaker in and of itself, even if you use DHCP, as many cable
and DSL setups that use DHCP in fact change the IP address very rarely.
One could even cobble together a script that determines the current "real"
hostname, rewrites main.cf to reflect the change, and reloads Postfix, and
then set that script to run after every IP address change (both dhcpcd and
dhclient can be configured for this, and if you use a router, you could
instead run the script as a cron job to test for such a change, then do
its thing if one has occurred). If your IP address changes often, that
hack might allow you to still use your system's "real" name in main.cf.
Note: if you change "myhostname =" in main.cf, be sure to append the
string ", earth.$mydomain" to the "mydestination =" line, so that Postfix
continues to be aware that the box sometimes goes by that name as well.
> Here was my config.php:
>
> $useSendmail = false;
> $smtpServerAddress = '192.168.0.3';
> $smtpPort = 25;
> $sendmail_path = '/usr/sbin/sendmail';
> $use_authenticated_smtp = false;
>
> I changed the ip address to 'localhost' and I haven't changed the
> 'useSendmail' option.
No problem, it's just talking SMTP directly to port 25, rather than
invoking the sendmail pseudo-app. No need to change anything else here.
> Yes, the postfix server and the squirrel server reside on the same box
> (as does most everything else).
Okay, and I gather that the Evolution box is a different one, but also on
the same LAN with the server system.
> Here are the relevant headers:
>
> Received: from 4.35.151.34 (EHLO servername) (4.35.151.34) by
> mta130.mail.sc5.yahoo.com with SMTP; Wed, 12 Nov 2003 12:14:06 -0800
> Received: from www.holt-tech.net (unknown
> [server.internal.ip.address]) by servername (Postfix) with SMTP id
> 13833205CFC for <[EMAIL PROTECTED]>; Wed, 12 Nov 2003 15:16:15 -0500
> (EST)
> Received: from evrtwa1-ar17-4-35-151-34.evrtwa1.dsl-verizon.net
> ([4.35.151.34]) (SquirrelMail authenticated user michael) by
> server.internal.ip.address with HTTP; Wed, 12 Nov 2003 12:16:15 -0800
> (PST)
>
> ************************************************************************
>
> Received: from 4.35.151.34 (EHLO servername) (4.35.151.34) by
> mta156.mail.scd.yahoo.com with SMTP; Wed, 12 Nov 2003 12:00:02 -0800
> Received: from machinename (unknown [host.internal.ip]) by
> servername (Postfix) with ESMTP id 0606E205CFC for
> <[EMAIL PROTECTED]>; Wed, 12 Nov 2003 15:02:11 -0500 (EST)
You should be able to get rid of the "unknown" bit in the latter set of
headers by putting an entry into the /var/spool/postfix/etc/hosts file on
the Postfix box that identifies the Evo system ("machinename") by tying
its internal IP address to its hostname. It would need to be here, as
Postfix runs chrooted (in its default MDK configuration), and cannot see
the "real" /etc/hosts file. You might also want to throw one in that pairs
the server's internal IP with "www.holt-tech.net"; that ought to clear up
the "unknown" in the former set of headers.
Still and all, the headers that pertain to the transactions between your
Postfix and Yahoo's SMTP server are essentially identical in both of the
examples, so it still escapes me why any destination system would be
willing to accept one variety of message and not the other. :(
Unless, of course, the only one giving you fits is your boss', which we
have already established is hosed in some bizarre fashion <g> ... but
having Postfix use a more valid hostname may fix that situation, too, even
though that doesn't fully explain that server's rather eccentric behavior.
HTH!
--
Bill Mullen [EMAIL PROTECTED] MA, USA RLU #270075 MDK 8.1 & 9.0
"In the beginning the Universe was created. This has made a lot of people
very angry and been widely regarded as a bad move." - Douglas Adams
Want to buy your Pack or Services from MandrakeSoft?
Go to http://www.mandrakestore.com
